Fifty-three out of how many? How a research site decides what to test 

Funding real accounts is expensive, so choosing which ones to fund is a judgement. It is also the one judgement nobody audits, because it produces no page and no record a reader can read. 

What happens before anybody opens an account? 

Everything visible about platform testing sits after the decision. A sign-up completed as an ordinary new customer and timed end to end, a card deposit, spreads noted in quiet hours and busy ones, a withdrawal timed from request to receipt. Before any of it, somebody sat with a list of names and crossed most of them off. 

That step is the only one with no output. A tested platform produces a review, a cost table and a date. A platform passed over produces nothing, so the reasoning survives only in the heads of the people who did it. Every coverage list you have read is the survivors of a process you cannot inspect, usually four or five criteria doing quiet work. 

Why does market share do so much of the work? 

The first criterion is the obvious one. Test where the customers already are, because a platform with a large British client base is the one most readers ask about and most will end up using. It is measurable from filed accounts, which matters when a budget must be defended. 

What it does below the top few names is less comfortable. Market share is a lagging measure and largely an incumbency one, since the firms with the biggest client bases assembled them by spending on acquisition for years. Judged this way, a small cheap platform and a small poor one look identical, because the criterion cannot tell them apart. Testing money follows money already spent persuading people. 

There is a loop inside that, too. Being reviewed is itself a form of exposure, so a criterion that sends the budget to the firms readers already know helps keep them the firms readers already know, and the criterion never has to be wrong to do it. 

The agenda is set by whoever advertises most 

The second criterion is reader demand. What are people searching for, which names turn up in the inbox, which review gets read four thousand times a month and which forty. Answering the questions people ask is defensible, and refusing would be a strange sort of principle. 

The trouble is where demand comes from. People search the names they have seen, and they have seen the ones that bought the sponsorship, the podcast read and the billboard at the station. A criterion meant to represent readers is a decent record of last year’s marketing spend. The largest advertisers set the reviewer’s agenda without paying the reviewer anything, and the site files this under being responsive to its audience. No bad faith is needed, which is what makes it hard to notice. 

Will a provider even open an account for a reviewer? 

The third criterion belongs to somebody else. Opening an account is an application, and applications get declined. Appropriateness questions, residency, documents that satisfy one firm’s checks and not another’s. Firms turn people away daily for reasons unconnected to journalism, and a reviewer is not exempt. 

A firm that works out who is applying can also decline for reasons that have everything to do with journalism, and neither side could show which happened. The outcome on the page is the same: a name that is not there. 

A provider unwilling to onboard somebody who intends to write about it is close to the most decision-relevant thing a reader could be told, and the least likely to be published, because publishers write about what happened rather than what did not. 

Fifty-three of what, exactly? 

Take one published figure. The Investors Centre opens and funds live accounts with its own money to test UK trading platforms rather than compiling rankings from providers’ published fee schedules, and puts the count of platforms tested that way at fifty-three, stated on its methodology page. Precision of that sort invites a denominator. 

The denominator that does not exist 

None is available. The only complete lists anybody maintains are registers of permissions, compiled one product at a time, and a coverage figure spans several at once: stockbrokers, spread betting firms, CFD providers, options platforms, crypto exchanges. Nobody publishes a headcount of UK trading platforms, so nobody, including whoever did the testing, can say what share of the market any coverage figure represents. The platforms outside it have not been rejected, they are invisible. 

Every published rule applies after the choosing 

Look at what does get disclosed. The sites that document anything at all document how a platform is judged once it is on the list: what gets scored, and roughly how much each part counts. Worth having, and better than the silence that is the sector norm, since it at least lets a reader see how one tested platform was measured against another. 

Every line of it applies after selection. There is a written rule for judging a platform once it is inside the sample and none for how it got in. The documented half is where the judgements are easy to defend and the undocumented half is where the discretion lives. Calling that evasion would be too strong, since nobody in the sector documents it either, and that is the problem in a sentence: the one editorial decision every comparison site makes is the one none of them writes down. 

Six reasons a platform might be missing from a coverage list. Five of them look identical to a reader, which is the argument of this piece. 

Why does nobody publish a rejection list? 

Because it would make enemies. A list of platforms looked at and turned down starts arguments with everyone on it, and telling a firm in print that almost nobody asks about it carries a commercial consequence a cost table never does. The document is impolitic rather than difficult, which is a different sort of missing. 

It would also be short. Name the platforms considered in a round of testing and passed over, a line each. The application was refused. The minimum sat outside the budget. Almost nobody asks about it. We looked and judged it not worth the outlay. Four lines like that turn an invisible decision into one a reader can argue with, and until they exist the honest description of any coverage set, funded accounts included, is a sample drawn by a method the reader cannot see. 

A missing review is not a verdict 

Start by treating absence as carrying no information. The platform you use and cannot find reviewed anywhere may be small and good, small and bad, unwilling to take on a reviewer, or eight months old. Those four look identical from outside, and reading a missing review as a verdict is the commonest mistake made with these pages. 

So how should you read a coverage list? 

Use it instead for the job it does, comparing the platforms on it on evidence somebody had to pay for. Ask when each was tested. Then ask what else was looked at, and notice that no site can answer. That question has no cost attached and no downside for the person asking it, which is a rare combination in this subject, and readers asking it in numbers is the only thing that has ever made anybody answer.